Multimodal treatment, including neoadjuvant chemotherapy (NACT) with or without radiotherapy (NACRT) followed by curative surgery, has become the standard of care for locally advanced esophageal squamous cell carcinoma (ESCC). This approach can downstage tumors, increase the likelihood of complete resection, and improve survival outcomes (1–4). Despite the uncertain survival benefits, traditional neoadjuvant therapy combined with programmed death-1 (PD-1) or programmed death-ligand 1 (PD-L1) inhibitors (neoadjuvant chemoimmunotherapy, NACI) has shown significant antitumor effects in recent years. This combination has led to a considerable pathological complete response (pCR) and acceptable toxicity, providing a promising alternative for treating this challenging disease (5–8). However, improved pCR rates have not consistently translated into improved survival benefits, as evidenced by results from several prospective randomized clinical trials (RCTs) (2, 6). It has been reported that recurrence occurs in 10–30% of those who achieve pCR (9–11). In a review of 3154 ESCC patients treated with NACT followed by surgery, Kitagawa et al. highlighted a strong correlation between recurrence-free survival (RFS) and overall survival (OS), underscoring the importance of addressing recurrence and the urgent need to explore alternative strategies to address this challenge.

Achieving complete resection of the esophagus with negative margins (R0 resection) is essential for esophagectomy. The margins encompass the proximal, distal, and circumferential resection margins (CRMs). While a 5-cm longitudinal margin is generally considered safe, the definition and implications of a positive CRM remain uncertain and debated (12, 13). In the context of neoadjuvant therapy, tumor regression often results in the formation of a scar or mass that can adhere tightly to surrounding tissues. In cases with random regression patterns (14–16), scattered tumor cells may remain in the mediastinum, potentially leading to locoregional recurrence (LRR), even when a negative CRM is reported.

Previously, we proposed a novel resection status, Rbr^+/-, which complements the R0 status of the Royal College of Pathologists (RCP) standards in patients after neoadjuvant therapy (7). In this pilot prospective study, we assessed the efficacy of Rbr^+/- in predicting LRR in patients with borderline resectable ESCC (Br-ESCC) (7). Preliminary results revealed that none of the patients (0/7) who underwent Rbr^- resection experienced LRR, suggesting that this grading system may serve as a potential predictor for recurrence and could guide postoperative treatment (7). Here, we describe a retrospective study aimed at exploring the relevance of Rbr^+/- and LRR and providing a foundation for future research, including our registered prospective clinical trial aimed at exploring the value of adjuvant radiotherapy in patients with Rbr⁺ status (ChiCTR2400082101).

This observational study presents a resection status, Rbr^+/-, in ESCC patients following neoadjuvant therapy, building on previously mentioned concepts (7). It distinguishes between Rbr⁺ and Rbr^- categories as a supplement to the R0 classifications defined by the RCP standard. Although similar results were preliminarily reported in our previous study (7), the current findings demonstrate for the first time that patients categorized as Rbr⁺ exhibit a significantly higher LRR rate (32 [22.7%] vs. 23 [7.6%], p<0.001, SMD = 0.430). OW is a PS method designed to mimic key characteristics of RCTs (17). Even after adjusting for imbalanced variables via OW, the LRR remained significantly higher in the Rbr⁺ group (20.1% vs. 11.4%, p=0.034, SMD = 0.242). The multivariable logistic regression model indicated that Rbr⁺ was an independent risk factor for LRR (p=0.018, OR: 2.19, 95% CI: 1.14-4.17). Additionally, the multivariate Cox analysis revealed that Rbr⁺ was associated with poorer LRRFS (p=0.008, HR: 2.56, 95% CI: 1.28-5.11) than Rbr^-. These findings suggest that Rbr⁺ might serve as a potential predictor of LRR following neoadjuvant therapy and surgery; thus, more intensive monitoring and appropriate adjuvant therapy are promising for patients with Rbr⁺ to improve their survival outcomes.

The mechanism by which Rbr⁺ status is associated with increased LRR despite complete resection (R0) remains unclear. We speculate that random tumor regression may be a contributing factor. Recently, the pattern of tumor regression was assessed in light of a proposed strategy for esophageal preservation following neoadjuvant therapy (14–16), although the findings remain to be interpreted and discussed. In 2013, Lanschot et al. reviewed 71 (70%) esophageal cancer patients whose residual tumor cells remained after NACRT (15). Despite the small proportion, the surrounding stroma with residual tumors accounted for 42% of cases, representing 30 patients (15). Even considering the regressive changes and/or residual tumor invasion into the surrounding stroma (ycT3), which pertains to patients initially involving all layers of the esophageal wall, 21% of the 51 overall patients exhibited a random regression pattern (15). Chuang and colleagues evaluated 76 patients with near pCR and found no rule of tumor regression (14). In addition, their results were inconsistent with those reported by Shapiro and colleagues (14). In contrast to these two studies (14, 15), Wang et al. provided a detailed description of regression patterns specifically in ESCC patients after NACRT. Four types of regression patterns have been proposed, with Type 4 (random regression) showing comparability to Types 1–3 (nonrandom regression) (16), which means that predicting tumor regression patterns remains challenging. On the basis of these results, when Rbr⁺ occurs, residual masses or proliferative tissue may persist in the mediastinum, as we previously demonstrated (7), whereas the resection margin is located in the tumor regression zone, resulting in pathological R0 and potentially contributing to development recurrence.

Generally, DM is reported as the predominant recurrence type of esophageal cancer following neoadjuvant therapy (18–20). However, in this study, LRR remained the primary type of recurrence compared with DM in the Rbr⁺ group ( Table 2 ). The significantly advanced ypT stage partially contributed to this outcome. Similar results have been reported in previous studies (7, 21). BRES-1 was our phase 2 study in which the safety and efficacy of NACI were assessed for patients with borderline resectable ESCC (7). Among the patients who underwent R0 resection (18 patients, 81.8%), four experienced LRR, whereas none had DM. When the patients were categorized as Rbr⁺ or Rbr^-, all recurrences were in the Rbr⁺ group (7). COSMOS is another multicenter phase 2 trial in which NACT was evaluated for initially unresectable locally advanced ESCC (21). After 3 years of follow-up, 7 (37%) of the 19 R0 patients had a recurrence or disease progression, while LRRs (4, 21%) slightly exceeded DMs (3,16%). Notably, in that study, the recurrence patterns in the non-R0 group were also reported, and a total of 29 patients (59%) were noted to have experienced tumor progression. In this group, the rate of LRR was higher than that of DM was (38% vs. 21%), further supporting the association between T stage and LRR (21).

In our study, Rbr status did not influence OS. When various AT regimens were compared with regard to OS in the Rbr⁺ group, the KM curves revealed that the median OS for the group receiving radiotherapy had not yet been reached. These findings suggest that appropriate AT might be a promising option for improving OS. Evidence is available from the JCOG1109 NExT clinical trial (2). In this study, LRR was more common in the NeoCF group (38 [38%] of 101) and the NeoCF+D group (33 [43%] of 76) than in the NeoCF+RT group (17 [23%] of 75 patients). Accordingly, radiotherapy or chemoradiotherapy was administered more frequently in the NeoCF and NeoCF+D groups, potentially contributing to improved OS (2).

Given the relatively high recurrence rate after neoadjuvant therapy plus surgery, AT is theoretically necessary. However, optimal patient selection, timing and regimen remain unclear, and previous studies have been inconclusive (22–25). In our cohort, Rbr status itself did not translate into inferior OS ( Table 4 , Figure 3B ), suggesting that appropriately tailored AT may mitigate the adverse impact of residual disease. Because LRR is more common among Rbr⁺ patients, adjuvant radiotherapy, delivered specifically to the mediastinum and tumor bed, could confer an OS benefit (26, 27). Indeed, we observed a trend toward improved OS in Rbr⁺ patients who received postoperative radiotherapy, although statistical significance was not reached ( Figure 3D ), likely due to the small sample size. Intriguingly, the “other” adjuvant category (chemotherapy alone or combined chemotherapy and immunotherapy) was independently associated with better OS in multivariable Cox analysis, further supporting the notion that adequate AT is critical in this subgroup. However, the trade−off between treatment−related toxicity and long−term benefit demands careful consideration, and the optimal AT regimen has yet to be established. Specifically, the integration of immunotherapy holds considerable promise for reshaping the therapeutic paradigm in ESCC in recent years (28–30). Prospective, large−scale trials are therefore essential to confirm the survival advantages of adjuvant radiotherapy, chemoradiotherapy, and chemo−immunotherapy in Rbr⁺ ESCC.

As a supplementary indicator to standard R0 resection, Rbr⁺ status was associated with a higher incidence of LRR, suggesting that such patients may represent a high-risk subgroup. While current consensus clearly recommends adjuvant therapy for patients with R1 or R2 resection, our findings imply that Rbr⁺ patients, although pathologically R0, might also benefit from closer surveillance and, in selected cases, consideration of adjuvant radiotherapy or intensified systemic therapy. This potential framework positions Rbr as a clinical tool to refine risk stratification and guide postoperative management, pending validation in prospective multicenter studies. Meanwhile, it must be emphasized that histopathological examination remains the gold standard for margin assessment. Rbr is not intended to replace pathology but rather to identify cases warranting heightened pathological scrutiny. We acknowledge that conventional block-based pathology may miss microscopic disease due to sampling limitations, which likely explains instances where intraoperative fibrosis (Rbr⁺) predicts recurrence despite negative margins. Future efforts should therefore also focus on refining pathological protocols, for example through systematic whole-mount sectioning, serial deeper sections, or adjunctive immunohistochemistry, to better capture subclinical tumor extension. Ultimately, bridging this gap with advanced pathology techniques will be essential for more definitive and reproducible margin evaluation.

To maximize the utility of Rbr as a supplementary tool and mitigate its inherent reliance on operator experience and subjective interpretation, future efforts must also focus on standardizing the Rbr assessment itself. Going forward, we plan to create a weighted Rbr index based on quantifiable imaging and intraoperative metrics. For example, we will measure the two−dimensional area of residual mass or fibrosis and assign incremental points for increasing size. Likewise, intraoperative dissection difficulty will be graded on a three−level scale (easy, moderate, difficult) with corresponding score values. In a prospective, multicenter setting, these component weights and threshold cut−points will be optimized and potentially adjusted using the four established pathologic regression patterns as modifiers. Such a structured algorithm will both standardize Rbr assessment across centers and allow automated or semi−automated implementation in future trials.

This study also has several limitations. Firstly, while we adjusted for several important confounders, this study lacked data on other potential variables since the characteristics of retrospective study. Secondly, information bias may be present; for example, patients who experienced recurrence or death during follow-up were more likely to be lost to follow-up or to provide inaccurate endpoint information. Thirdly, it remains uncertain whether the Rbr^+/- system is applicable to patients undergoing NACRT, as most participants in this study received NACT. Fourthly, the number of patients who received radiotherapy was relatively small, which warrants confirmation in larger-scale studies. Finally, as a single-center retrospective study, our findings require validation through prospective multicenter trials before clinical implementation.